Crossover designs

These are designs for experiments in which treatments are applied to several people or animals (or other units) and the treatments applied to an individual can be changed from one period to the next. What makes these different from ordinary row-column designs is that the treatment applied in the preceding period may have a carryover effect on the person or animal in the current period.

Such experiments are used to compare drugs to alleviate symptoms of chronic disease; to compare feeds for lactating cows; to compare the tastes of different cheeses (or wines or beers or types of orange juice …) or find out about human-computer interaction.

Topics	Some of my publications
How should we design these experiments when we want to compare the estimated effects of giving each treatment long-term?	R. A. Bailey and P. Druilhet: Optimality of neighbor-balanced designs for total effects. Annals of Statistics 32 (2004), 1650–1661. doi: 10.1214/009053604000000481 [Maths Reviews 2089136 (2005h: 62191)]
How should we design these experiments when it is assumed that the carryover effect of each treatment is proportional to the direct effect of that treatment?	R. A. Bailey and J. Kunert: On optimal crossover designs when carryover effects are proportional to direct effects. Biometrika 93 (2006), 613–625; doi: 10.1093/biomet/93.3.613 [Maths Reviews 2661446]
How should we design these experiments when it is assumed that there is full interaction between the direct effects and the carryover effects?	R. A. Bailey and P. Druilhet: Optimal cross-over designs for full interaction models. Annals of Statistics, 42 (2014), 2282–2300. doi: 10.1214/14-AOS1247

Page maintained by R. A. Bailey